The Regress Argument Against the Language of Thought Hypothesis

This objection has plagued the Language of Thought hypothesis (aka LOTH) since it's very conception. In fact Jerry Fodor included a reply to such objections in his 1975 book the Language of Thought, however these objections have refused to go away. It dubious whether such objections to the Language of Thought hypothesis do in fact under the hypothesis. 

Many of the supporters of LOTH often appeal to the idea of a language of thought to explain a number of things about our natural languages. For example the language of thought has been appealed to in order to explain the following: 

• How natural languages (i.e French, English, Russian etc.) are learned 
• How our natural languages are to be understood 
• And finally how utterances in a natural language can be meaningful 

There are couple of instances in Fodor (1975) which empathize the points above. For example Fodor claims that natural languages are learned through a process of forming and confirming hypothesis regarding the translation of sentences in natural language into mentalese sentences. The LOTH gives us a representational medium in which the representation of the truth conditions of natural language sentences can occur. The Language of Thought theorist further posits that our natural languages are understood because such understanding consists in the translation of our standard natural language sentences into sentences in mentalese. Again sentences in our standard natural language are taken to meaningful in virtue of the meanings of the corresponding mentalese sentences.  

Simon Blackburn (1984) has claimed that either explanations of this kind lead to an regress as we ought to have to give an explanation of how a language of thought is learned, or that they are simply gratuitous because if it is possible to give a successful explanation which doesn't lead to regress, we ought to have given such an explanation for our natural languages without positing a language of thought at all. 

Fodor's response to such objections can be found in Fodor (1975) and goes roughly as follows:

• The Language of Thought is innate it is not learned.
• The Language of Thought is understood in a different sense to how we understand our natural language comprehension. 
• That sentences in a language of thought are not meaningful in virtue of another meaningful language but rather in a completely different way. 

While Fodor's replies to the regress argument aren't totally plausible and may not convince everyone. But Laurence and Margolis (1997) have pointed out that the regress argument against LOTH (at least in the form that Blackburn presents it) relies on the assumption that the language of thought is only posited to explain certain facts about natural language acquisition and natural language comprehension. If it can be shown that there is other empirical phenomena for which the LOTH provides a good explanation then the hypothesis is not gratuitous in the way Blackburn suggests. In fact there does seem to be a good deal empirical evidence that can be best explained by positing a LOT (consider both the systematicity and productivity argument). 

Dennett's Objection to the Language of Thought Hypothesis

The well known philosopher Daniel Dennett formulated one of the first serious objections to the language of thought hypothesis (LOTH) in his review to Jerry Fodor's (1975) foundational book on the subject. Dennett's objection is essentially that it is possible for their to be propositional attitudes without explicit representation. To make his point Dennett introduces an example:

In a recent conversation with the designer of a chess-playing program I heard the following criticism of a rival program: “it thinks it should get its queen out early.” This ascribes a propositional attitude to the program in a very useful and predictive way, for as the designer went on to say, one can usefully count on chasing that queen around the board. But for all the many levels of explicit representation to be found in that program, nowhere is anything roughly synonymous with “I should get my queen out early” explicitly tokened. The level of analysis to which the designer's remark belongs describes features of the program that are, in an entirely innocent way, emergent properties of the computational processes that have “engineering reality.” I see no reason to believe that the relation between belief-talk and psychological talk will be any more direct. (Dennett 1981: 107) 

The rival but critical chess programmer assigns a propositional attitude to the rival program; namely that the rival program thinks it should move its queen out into play early in the game. Such an ascription of a propositional attitude is both useful and predictive. For example when we wish to program our chess computer to play the rival program we may consider the fact that the other program thinks it should get its queen early and respond by making sure we have an adequate defense for such an event. But if we understand how chess programs operate we know that there is no internal representation within the code of a chess computer which represents the propositional attitude that the program 'should get its queen out early'. Dennett sees no reason why the relation between our standard everyday belief talk and talk of psychological processes will be anymore direct than in the chess program/computer example. 

The chess program doesn't have  a dispositional or potential, belief regarding the early movement of its queen. Rather it operates with the belief that it ought to get its queen out early in the game.There appears to be lots of everyday examples where we reason using certain rules of inference without directly or explicitly representing this rules of inference.  

This objection from Dennett hasn't been particularly well received and it is widely regarded that Language of Thought theorists can provide a more than an adequate reply to such objections. The standard reply involves distinguishing between the rules regarding the way Mentalese data structures are manipulated and the data structures themselves. The Language of Thought hypothesis is not committed to every rule being explicitly represented. It is a nomological fact that in a computational device can be explicitly represented some have to be hard wired into the system. It is in this way that language of thought theorists do not have to contend that rules will be explicitly represented, it is data structures that have to be explicitly represented. With these data structures being manipulated formally by rules, causal manipulation is not itself possible without explicit tokening of these representations. It is possible to account for dispositional propositional attitudes in terms of an appropriate principle of inferential closure of explicitly represented propositional attitudes. 

A chess program involves at least some certain explicit representations (for example chess board, pieces and some of the rules). Which of the rules of the program are explicit and which are implicit is an empirical matter of fact. What Dennett's objection essentially does it point to the fact that some rules may be emergent out of the implementation of explicit rules and data structures. This does not undermine the language of thought hypothesis, as it possible to account for these emergent rules in terms of data structures and explicit representations. 

References 
Dennett D (1981), Brainstorms: Philosophical Essays on Mind and Psychology, Cambridge, Massachusetts: MIT Press, 1981